Danny Sullivan, one of Google’s SEO experts, has published an article that explains how SEO works. Auto-completion predictions from the google search page.
These are the factors that Google uses to make the auto-complete predictions.
To find related results and predictions, Google uses information about the most common searches and trending queries made by users who have already done a similar search. In addition, the results are selected based on certain criteria such as the user’s language (the one configured in their browser), the language in which the search is carried out and the latest news.
Among the factors that Google uses for its predictions are the views and searches of users on YouTube, even though the video platform has its own prediction system. This is how Danny Sullivan explains this point on his Twitter account:
Auto-complete search isn’t pulled from your YouTube searches to my knowledge. YouTube has its own autocomplete system.
– Danny Sullivan (@dannysullivan) October 8, 2020
Google has a system in place that prevents its algorithm from starting predictions of little value or results Failure to comply with quality guidelines from Google. Unless a specific search is performed, such as sexually explicit content, Google’s auto-complete predictions will never show violent, hateful, dangerous, or sexually explicit content.
In September, Google updated its quality guidelines for auto-complete predictions from Google. From then on, all election content or calls to vote for certain parties are excluded. In addition, it uses the new BERT algorithm to find content syntactically related to search queries, while discovering and displaying new messages in a few minutes than you previously had to wait around 40 minutes.
In 2018, Google already made some changes to its quality policy for its predictions. On that occasion, the company focused on eliminating content that could be viewed as spam or a source of fake news. Additionally, since 2017, Google has had a system for reporting results that are offensive or that do not meet the company’s quality standards.